Computer networks facilitate communication between computers, servers, and stand-alone peripherals. They make possible large scale computing systems, distributed service systems, and a whole host of applications that would otherwise be infeasible. Therefore, incentives exist for the use and enhancement of computer network technology.
Storage area networks (SANs) are a popular type of computer network for accessing large volumes of data. A SAN is a networked infrastructure connecting servers to stand-alone data storage devices (e.g. disk drives) over a high-speed network. The SAN is usually a sub-network of a larger computer network that include servers and personal computers that need to access the data in the SAN. The advantages of SANs include having a large, shared storage capacity that has high bandwidth access and does not have to be accessed through a single source.
Many SANs rely on the Fibre Channel (FC) protocol. A single FC link can carry data at rates exceeding 2 gigabits per second (Gb/s) in both directions simultaneously. The FC protocol defines standard media and signaling conventions for transporting data in a serial fashion. It also provides an error correcting channel code and a frame structure for transporting the data. Further, the FC protocol sets out a buffer-credit-based flow control methodology, and creates some common services to allow proper routing of data (e.g. fabric controller, name server). The FC protocol can be applied to various network topologies including point-to-point, ring, and switched fabric.
Many FC switches provide at least some degree of automatic configurability. For example, they may automatically sense when a new inter-switch link (ISL) becomes active, and may initiate an initialization process to discover what the link connects to. The switch may automatically determine various parameters for the link (e.g. link speed). As FC networks are created, updated, maintained and de-commissioned, switches may be enabled, disabled or reconfigured, and links may be added or removed.
Routing of information between the switches will change accordingly due to a routing protocol called Fibre Channel Shortest Path First (FSPF). This protocol uses information about the “cost” of all ISLs in the network, referred to as a topology database. The cost may represent an integer greater than zero. The FSPF protocol uses the topology database to compute a routing table associated with a respective switch. This routing table exists in each switch and contains the output port a particular frame may exit the respective switch on. The routing table does not contain information regarding other switches in the fabric. When an individual switch receives a frame, the routing table is utilized to determine what port to send the frame out upon.
The time it takes a frame to traverse from its source to destination in the network is referred to as the latency of the route. In multi-switch networks, frames can be routed through numerous switches before arriving at their destination. Each switch constitutes a hop that typically adds 1 microsecond or more of latency to the route. Congested or oversubscribed routes in large networks may have latencies of more than fourteen microseconds. Congestion and over subscription of ISLs may lead to significant performance problems due to increased latency.
Since routing within the Fibre Channel has some degree of automatic configurabilty, the actual route a particular frame takes to reach its destination becomes highly variable in large networks due to switch failures and the activation of new ISLs. This variability may be complicated by the use of multiple, parallel ISLs between switches to form ultra-high bandwidth “trunks”. In addition, the FSPF protocol permits load sharing among multiple, equal-cost paths. As such, multiple paths may be utilized to balance the traffic among these paths.
FC networks can grow quite large. The protocol allows for nearly 224 (over 16 million) node ports within a single fabric (an FC network includes one or more FC fabrics). Each node port supports one FC device. As larger networks are implemented (e.g. more than about seven switches), troubleshooting performance drops becomes a daunting task. For example, ISLs and ports may become congested along particular routes, significantly reducing performance. It would be desirable to identify routing issues as a preliminary step to eliminating or mitigating the adverse effects, thereby improving the speed, efficiency, and reliability of larger networks.
TCP/IP networks often use a utility called traceroute to show the path of a packet. Basically, traceroute works by sending a series of ICMP echo packets, with the allowed hop count increasing by one for each packet. When the hop count is exceeded, the router returns an expired message. By collecting these expired messages and the ultimate echo message, in combination with the incrementing hop count, the path to the desired host can be determined. However, that and the time of each hop are all that are known. This minimal amount of information would not really assist in troubleshooting a FC SAN. Further, because traceroute operates by sending multiple, time-delayed packets, the returned path may actually be erroneous should the routing change in between packets. This further limits the potential usage of a traceroute approach in a FC SAN environment. It would be desirable to provide a mechanism to obtain greater information and avoid other problems inherent in solutions such as traceroute.
Token ring networks also include a route discovery technique that must be used if a frame must traverse multiple networks. To determine a route to an unknown destination, a source device provides a route discovery or explorer frame. This frame is fanned out to every ring in the LAN segment by the interconnecting bridges. As the frame is forwarded from one ring to the next, the bridge updates routing information in the route discovery frame by including the ring ID and bridge ID in the frame. When the frame eventually reaches the destination, the full route is contained in the frame. The destination uses this information to develop the source routing information used to provide a response frame to the source device. This source routing information is then included in every frame that goes between the two networks. While this technique addresses some of the problems of the traceroute approach, it still only provides minimal information and would be of minimal use in a FC environment. Further, in an FC environment all routes are already known due to the FSPF routing protocol, so this token ring technique would be superfluous in an FC environment.